MeshAnything: 3D Generator — What It Does and What You Need to Run It

MeshAnything is a 3D generator that turns input data into polygon meshes. The authors position it as a tool capable of producing meshes that resemble the work of modeler artists, using autoregressive transformers. The project was presented at the ICLR 2025 conference.

What it does

• Generates 3D meshes from input data — the authors state that the system works on the principle of "from anything to mesh."
• Uses an autoregressive model based on the transformer architecture.
• As described by the developers, creating a single mesh takes about 30 seconds on an A6000 accelerator.

What you need to run it

• Platform: as described by the authors, the environment was tested on Ubuntu 22 with CUDA 11.8. CUDA support is required to run it.
• VRAM: the authors state that generating a mesh requires about 7 GB of VRAM on an A6000 accelerator.
• Language: Python.
• License: S-Lab License 1.0: non-commercial use only.
• Code: open source, available at github.com/buaacyw/MeshAnything. The last change is dated April 28, 2025.

Who it suits

• Researchers in generative 3D and machine learning who want to study or refine the autoregressive approach to mesh creation.
• Developers ready to run the model on machines with CUDA-compatible accelerators and at least 7 GB of VRAM.
• Those looking for open source for mesh generation but willing to accept the non-commercial use restriction.

The tool is not suitable for users without access to discrete NVIDIA GPUs with CUDA support, nor for those expecting a ready-made application with a graphical interface — the authors provide only Python code.

MeshAnything is an open-source research project that offers one possible approach to polygon mesh generation. Before using it, you should carefully review the license terms, as they differ from standard open licenses.